KR102229013B1 - Automatic stream detection and assignment algorithm - Google Patents

Abstract

Solid state drive is listed. The solid state drive includes a flash memory to store data and can support multiple device streams. The solid state drive controller can read or write data to flash memory, and can store submission queues and chunk-stream mappers. The flash conversion layer includes a receiver that receives a write command, a logical block address mapper that maps a logical block address to a chunk identifier, a stream selection logic that selects a stream identifier based on the chunk identifier, and a stream identifier that adds a stream identifier to the write command. It may include an adder, a liner to place the chunk identifier in the submission queue, and background logic to update the chunk-stream mapper after the chunk identifier has been removed from the submission queue.

Description

Automatic stream detection and allocation algorithm {AUTOMATIC STREAM DETECTION AND ASSIGNMENT ALGORITHM}

Technical concepts of the present invention relate to solid state drives (SSDs), and more particularly, to managing streams in multi-stream solid state drives.

Multi-streaming solid state drives (SSDs) allow smart placement of input data, minimizing the impact of internal garbage collection (GC) and reducing write amplification. Multi-streaming is achieved by adding a simple tag (stream identifier) to each of the write commands sent from the hose to the solid state drive. Based on this tag, the solid state drive can group data into common blocks.

However, in order to use multi-streaming devices, applications support solid state drive multi-streaming so that software sources can allocate common streams to data with similar properties such as data lifetime. You should know. Making software multi-stream-aware requires modification of the software. However, modifying any software carries the risk of making unintended changes to the behavior of that software. Given the numerous different software products on the market, expecting that a handful of these software products will be modified to support multi-streaming seems to be a unlikely proposal at best.

There is a need for a way to support multi-streaming without software requiring modifications.

An object of the present invention is to provide an automatic stream detection and allocation algorithm that allocates and manages a stream identifier by reflecting the access characteristics of a chunk.

An embodiment of the inventive concept includes a solid state drive (SSD). The solid state drive may include a flash memory for storing data; Support for a plurality of device streams in the solid state drive; A solid state controller including storage for a submission queue and a chunk-stream mapper, and managing writing data to the flash memory in response to a plurality of write commands; And it includes a flash conversion layer. The flash translation layer includes: a receiver for receiving a write command including a logical block address (LBA); A logical block address mapper for mapping the logical block address to a chunk identifier (ID); Stream selection logic for selecting a stream identifier based on the chunk identifier using the chunk-stream mapper; A stream identifier adder for adding the stream identifier to the write command; A line-up device for placing the chunk identifier in the submission queue; And background logic to lift the chunk identifier from the submission queue and update the chunk-stream mapper.

An embodiment of the inventive concept includes a driver for use in a computing system, wherein the driver is a write command for a solid state drive (SSD) including a logical block address (LBA). A receiver for receiving; A logical block address mapper for mapping the logical block exchange to a chunk identifier (ID); Stream selection logic for selecting a stream identifier based on the chunk identifier using a chunk-stream mapper stored in a memory of the host computer system; A stream identifier adder for adding the stream identifier to the write command; A line-up device for placing the chunk identifier in a submission queue stored in the memory; And background logic to remove the chunk identifier from the submission queue and update the chunk-stream mapper.

An embodiment of the technical idea of the present invention includes a method. The method includes receiving a write command from a software source; Determining a logical block address (LBA) in the write command; Identifying a chunk identifier (ID) for a chunk in a solid state drive (SSD) including the logical block address; Accessing a stream identifier associated with the chunk identifier; Allocating the stream identifier to the write command; Processing the write command in the solid state drive using the allocated stream identifier; And performing a background update of the stream identifier associated with the chunk identifier.

According to the present invention, a stream identifier is assigned in consideration of the sequence, frequency and latest. Further, according to the present invention, the stream identifier allocated to the chunk is adjusted in consideration of the access count and the expiration time. Accordingly, an automatic stream detection and allocation algorithm that allocates and manages a stream identifier by reflecting the access characteristics of a chunk is provided.

1 shows a machine having a solid state drive according to an embodiment of the inventive concept.
FIG. 2 shows additional details of the machine of FIG. 1.
3 shows details of the solid state drive of FIG. 1.
4 shows details of the flash transformation layer of FIG. 3.
FIG. 5 shows the logical block addresses of various instructions mapped with chunk identifiers and later mapped with the stream identifiers for the solid state drive of FIG. 1.
6 shows the various commands of FIG. 5 modified to include the stream identifiers of FIG. 5 and transmitted to the solid state drive of FIG. 1.
FIG. 7 shows an arithmetic logic unit that maps the logical block addresses of FIG. 5 with the chunk identifiers of FIG. 5.
8 shows a sequence, frequency, and latest table used to map chunk identifiers to stream identifiers according to a first embodiment of the technical idea of the present invention.
9 shows additional details of the flash transformation layer of FIG. 3 and the background logic of FIG. 4 according to a first embodiment of the inventive concept.
10 shows details of the sequential logic of FIG. 9.
Figure 11 shows the calculation of the latest weighting using the latest logic of FIG. 9.
FIG. 12 shows that the access count adjuster of FIG. 9 adjusts the access count using the latest weighting of FIG. 11.
FIG. 13 shows the calculation of a stream identifier from the adjusted access count of FIG. 12 using the stream identifier adjuster of FIG. 9.
14 shows a node that can be used to map chunk identifiers with stream identifiers according to a second embodiment of the inventive concept.
15 shows details of the background logic of FIG. 4 according to a second embodiment of the inventive concept.
16 shows promotion and demotion of chunk identifiers in the queue of FIG. 15.
17 shows details of the promotion logic of FIG. 15.
18 is a view illustrating calculating a stream identifier from an adjusted access count according to a second embodiment of the inventive concept.
19 shows details of the chunk expiration logic of FIG. 17.
20 shows details of the downgrade logic of FIG. 15.
21A and 21B are flowcharts of exemplary procedures for determining a stream identifier for a write command of FIG. 5 according to an embodiment of the inventive concept.
22 is a flowchart of an exemplary procedure for a logical block address mapper for mapping logical block addresses of FIG. 5 with chunk identifiers of FIG. 5 according to an embodiment of the inventive concept.
23A and 23B are flowcharts of an exemplary procedure for updating a stream identifier for a chunk using sequential logic according to a first embodiment of the inventive concept.
FIG. 24 is a flowchart of an exemplary procedure for performing a background update of the SFR table of FIG. 8 according to a first embodiment of the inventive concept.
25A to 25C are flowcharts of exemplary procedures for performing a background update of a node of FIG. 14 according to a second embodiment of the inventive concept.

Embodiments of the present invention will be described with reference to the accompanying drawings. In the detailed description that follows, many specific details are set forth for a thorough understanding of the invention. However, it will be understood that the present invention may be practiced by those skilled in the art without these specific details. In other words, well-known methods, procedures, components, circuits, and networks have not been described so as not to unnecessarily obscure the embodiments.

Although terms such as first and second are used to describe various constituent elements, it will be understood that constituent elements are not limited by these terms. These terms are only used to distinguish components from each other. For example, without deviating from the technical idea of the present invention, a first module may be referred to as a second module, and similarly, a first module may be referred to as a second module.

Terms used in the description of the technical idea of the present invention are for describing specific embodiments, and are not intended to limit the technical idea of the present invention. It will be understood that the term "and/or" as used herein refers to any or all possible combinations of one or more related items. The terms "comprising" and/or "comprising" describe the presence of the recited features, integers, steps, actions, and/or components, and one or more of the features, integers, and steps. Does not exclude the presence or addition of elements, actions, components, and/or combinations thereof. Components and characteristics of the drawings are not drawn to scale.

An apparatus and method for performing automatic stream detection and allocation independently of an application layer are proposed. Stream allocation is based on real-time workload detection and can be independent of application(s). Stream allocation can be applied to solid state drives (SSDs) available in multi-streams.]

Implementation of the stream allocation protocol has several advantages. First, the applications do not need to be modified at all. However, applications that assign their own stream priorities are considered in this attempt. For example, the stream allocation protocol is entirely dependent on application-allocation stream priority. Alternatively, the stream allocation protocol may perform a weighted sum combining the application-allocation stream priority and the calculated stream allocation using any necessary weights. Second, any application that does not use devices available in multi-stream can use automatic stream detection and allocation. Third, applications do not need to know the stream-associated information from the hardware. Fourth, the stream allocation protocol manages each multi-stream-active solid state drive separately, allowing applications to use multiple solid state drives, and multi-stream-active and non-multi-stream-active solid state drives. It also makes it possible to mix them. Fifth, stream allocation is done in the file system, in the block layer, in the device driver, inside the solid state drive, if the start address of the write command (or file offset, if the stream allocation protocol is implemented in the file system layer) can be identified. It can be performed in any required layer of the system, such as (eg, in the flash conversion layer).

The total address space of solid state drives can be divided into a number of fixed size chunks. This chunk size can be any desired size, for example a multiple of 512-byte sectors. The starting logical block address (LBA) of the request can be converted to a chunk identifier (ID) by dividing the logical block address by the number of sectors per chunk, more generally by the chunk size.

In the first embodiment of the technical idea of the present invention, sequential, frequency, and sequential, frequency, and recency (SFR) approaches may be used to determine stream allocation.

● Sequential: If the starting logical block address of the new request is adjacent to the ending logical block address of the previous requests (i.e., if write commands are associated with sequential writes), then the second write command will be assigned the same stream identifier as the previous write command. I can. This approach assumes that a group of sequential requests will have a similar lifetime if issued within a short period of time. Since write commands can be issued by different sources (applications, file systems, virtual machines, etc.) and can be mixed, write commands from a solid state drive's point of view do not follow a strict sequential pattern. May not. To handle this possibility, the sequence can be determined in association with the number of previous instructions in the window (the size can vary). For example, a window size of 4 will determine whether the incoming request is sequential with which of the previous four requests.

● Frequency: The frequency refers to the number of times the starting logical block address is accessed. The frequency can be measured in access counts. Each time a chunk is accessed (written), the access count for that chunk is incremented by one. Higher access counts indicate a shorter lifetime for that chunk. Thus, the frequency reflects the temperature of the data chunk.

• Latest: Latest refers to the temporary locality of data chunks. For example, one chunk is accessed more frequently during a certain period of time and accumulates a high access count, but may not be used afterward. In this situation, there is no need to keep the chunk in the hot stream for a long period of time. Chunks are considered hot only if they are accessed frequently within the most recent period. In an embodiment of the inventive concept, the decay period may be predefined for all chunks. If the chunk is not accessed within the last N decay periods, the access count can be divided by ^2N.

The stream identifier for the chunk can be determined by both frequency and latest (and sequential, if applicable). When chunks are accessed frequently, the frequency promotes the chunk to a hotter stream, while the latest demotes that chunk to a cooler stream if the chunk is not used during the last decay period. In an embodiment of the inventive concept, a log scale may be used to convert the access count into a stream identifier, which indicates that the number of device streams is much smaller than the range of values for the access counts. Can be reflected. So, for example:

Latest weight = 2 ^{(( current time -last access time) / (decay period),}

Stream identifier = log (access count / latest weight)

The update of the stream identifier for the chunk can be performed as background tasks to minimize the impact on input/output (I/O) performance.

In the second embodiment of the inventive concept, a multi-queue approach may be used. For each device stream, a queue can be defined. Since there are multiple device streams, multiple queues may exist. The multi-queue algorithm can be divided into two different functional modules. One module handles the promotion of each chunk from a lower queue to a higher queue, and the other module handles the demotion of less used or unused chunks from a higher queue to a lower queue. The higher the access count of a chunk, the hotter the chunk is considered to be, so that chunk is placed in a higher queue. Both promotion and demotion can be performed as background tasks, minimizing the impact on I/O performance.

When a chunk is first accessed, it is assigned to stream 0 (lowest stream) and placed in the corresponding queue. Otherwise, the chunk is removed from the current queue, the access count of the chunk is updated, and the chunk is placed in a new queue (potentially). Any attempt can be used to determine an appropriate queue based on the access count. For example, the log of the access count can be calculated with the new stream identifier (identifying the corresponding queue). The promote module can also check if the chunk is currently the hottest chunk (based on the access count). If the chunk is currently the hottest chunk, the device lifetime can be set based on the interval between accesses of this chunk. Finally, the promotion logic can determine the expiration time of the chunk based on the device lifetime and the last access time of the chunk.

To demote a chunk, the demote module can check if the chunks reach the head of the queue. If the chunk has not yet passed its expiration time, the chunk can be left alone. Otherwise, the chunk may be removed from the current queue, assigned a new expiration time, and demoted to a lower queue (ie, assigned a lower stream identifier).

Additionally, if the chunk being degraded was the hottest chunk, that chunk was not accessed for a while (as judged by the expiration time) (so that chunk is no longer hot, and another chunk from that queue is the hottest chunk). Can be selected (with an appropriate derivation of the device lifetime), the newly selected hottest chunk may be the next chunk in that queue, or the last chunk that entered that queue.

1 shows a machine having a solid state drive (SSD) according to an embodiment of the inventive concept. In FIG. 1, a machine 105 is shown. Machine 105 may be any necessary machine including, without limitation, a desktop or laptop computer, a server (standalone server or rack server), or any other device that may benefit from embodiments of the inventive concept. have. Machine 105 may also include specialized portable computer devices, tablet computers, smartphones, and other computing devices. Machine 105 can execute any necessary applications. Database applications can be good examples. However, embodiments of the technical idea of the present invention can be extended to any necessary application.

Regardless of the particular form, the machine 105 may include a processor 110, a memory 115, and a solid state drive (SSD) 120. The processor 110 may be any of a variety of processors, for example, an Intel Xeon, Itanium, or Atom processor, an AMD Opteron processor, an ARM processor, and the like. 1 shows a single processor, machine 105 may include any number of processors. The memory 115 may be a flash memory, a static random access memory (SRAM), a persistent random access memory, a ferroelectric random access memory (FRAM), or a magnetoresistive random access memory. Memory (MRAM, Magnetoresistive Random Access Memory), such as non-volatile random access memory (NVRAM, Non-Volatile Random Access Memory) can be any of a variety of memory, such as, typically DRAM. Memory 115 may also be any desired combination of different memory types. The memory 115 may also be controlled by a memory controller 125 that is part of the machine 105.

Solid state drive 120 may be any of a variety of solid state drives and may be expanded to include other types of storage that perform garbage collection (even when not using flash memory). The solid state drive 120 may be controlled by the device driver 130 present in the memory 115.

2 shows additional details of the machine 105 of FIG. 1. 2, typically, machine 105 includes a memory controller 125 and a clock 205, and one or more processors used to coordinate the operations of the components of machine 105. Includes 110. Processors 110 may also be coupled with memory 115, which may include, for example, random access memory (RAM), read-only memory (ROM), or other state storage medium. Processors 110 may also be coupled with solid state drives 120 and a network connector 210, which may be, for example, an Ethernet connector or a wireless connector. The processors 110 can also be attached to the user interface 220, among other components, and a bus that can be attached to input/output interface ports that can be managed using the input/output engine 225. 215).

3 shows details of the solid state drive 120 of FIG. 1. In FIG. 3, the solid state drive 120 includes a host interface logic 305, a solid state drive controller 310, and various flash memory chips 315-1 organized into various channels 320-1 to 320-4. ~315-8). Host interface logic 305 may manage communications between solid state drive 120 and machine 105 of FIG. 1. The solid state drive controller 310 may manage read and write operations along with garbage collection operations in the flash memory chips 315-1 to 315-8. The solid state drive controller 310 may include a flash translation layer 325 that performs some of this management. In embodiments of the inventive concept having a solid state drive 120 responsible for assigning write commands to streams, the solid state drive controller 310 includes a storage 330 that supports stream allocation. can do. Repository 330 may include a submission queue and a chunk-stream mapper 340. The submission queue 335 may be used to store information about chunks affected by various write commands. As write commands are received, chunks (or rather identifiers (IDs) of these chunks) associated with those write commands may be placed in the submission queue 335. Thereafter, (affects front operations) To reduce), as part of the background process, chunks are removed from the submission queue 335, and the stream assignment for these chunks can be updated. The chunk-stream mapper 340 allows some streams to be assigned to various chunks. It is possible to store information about whether it is currently allocated, as a result of chunk identifiers in the submission queue 335 (or as a result of a lack of chunk identifiers in the submission queue 335) (unused chunks are unused). May be assigned to lower priority streams as a result of the) The concept of chunks is further discussed with reference to Figs. 5 and 7 below.

In FIG. 3, the solid state drive 120 is shown to include eight flash memory chips 315-1 to 315-8 organized into four channels 320-1 to 320-4. Embodiments of the technical idea may support any number of flash memory chips organized into any number of channels.

4 shows details of the flash transform layer 325 of FIG. 3. In FIG. 4, the flash conversion layer 325 includes a receiver 405, a logical block address (LBA) mapper 410, a stream selection logic 415, a stream identifier adder 420, and a transmitter 425. ), a queue 430, and background logic 435. Receiver 405 may receive write commands from various software sources such as operating systems, applications, file systems, remote machines, and other such sources. For a given write command, the logical block address mapper 410 may map the logical block address used in the write command to a specific chunk of the solid state drive 120 of FIG. 1. The stream selection logic 415 may later select a stream suitable for the corresponding chunk. The stream selection logic 415 can achieve this stream selection using the chunk-stream mapper 340 of FIG. 3, and searches for the chunk-stream mapper 340 of FIG. 3 to obtain an entry corresponding to the selected chunk. You can include logic to find (entry). Otherwise, the stream selection logic 415 may, for example, calculate the stream identifier from the access count for the chunk (similar to that described with reference to Figure 13 below) or by allocating streams in a round robin access (write Other approaches can be used, distributing instructions evenly across all device streams. The stream identifier adder 420 may add the selected stream identifier to the write command by using logic for writing data to the write command. When the stream identifier is attached to the write command, the transmitter 425 may transmit the write command (along with the attached stream identifier) to the solid state drive 120 of FIG. 1 for execution.

The line up 430 may have the chunk identifier identified for the write command, and add the chunk identifier to the submission queue 335 of FIG. 3. The line up 430 may use logic to add the chunk identifier to the submission queue 335 of FIG. 3. For example, the line up 430 includes a pointer to the tail of the submission queue 335 of FIG. 3, and writes the chunk identifier to the tail of the submission queue 335 of FIG. I can. Thereafter, the pointer to the tail of the submission queue 335 of FIG. 3 may be updated to point to the next slot of the submission queue 335 of FIG. 3. Eventually, the chunk identifier will be released from the submission queue 335 of FIG. 3, after which background logic 435 may operate on the released chunk identifier. Background logic 435 is further discussed with reference to FIGS. 9 and 15 below.

Background logic 435 can be implemented in any desired way. For example, the background logic 435 may operate as simple as incrementing the stream identifier each time the chunk is accessed (this is the logic that performs a search in the data structure storing chunk identifiers and stream identifiers, and Including nothing more than an incrementer that increments the stream identifier located in the method). However, this simple implementation eventually means that all chunks will use the highest numbered stream and the lower numbered streams will not be used. More relevant implementations can consider if the chunk is not heavily used and lower the stream priority of the chunk to match. Two embodiments of the technical idea of the present invention implementing the background logic 435 in different ways to achieve this result will be described below with reference to FIGS. 8 to 13 and FIGS. 14 to 20.

Background logic 435 is described as operating in the background (and hence its name), but performing stream allocation updates in the background minimizes the impact on reads and writes to the solid state drive 120 of FIG. It is convenient to use. If the stream assignments can be updated without affecting the performance of the solid state drive 120 of FIG. 1, there is no reason for the background logic 435 to fail to operate in the foreground. For example, if the solid state drive 120 of FIG. 1 includes a processor (that is, if the solid state drive 120 of FIG. 1 proposes In-Storage Computing (ISC)), FIG. 1 This processor in the solid state drive 120 of FIG. 1 could potentially be used to update stream assignments in real time without affecting the read and write performance of the solid state drive 120 of FIG. 1. In this situation, the background logic 435 is immediately without having to place the chunk identifier in the submission queue 335 of FIG. 3 or wait when the background logic 435 can operate without affecting the foreground operations. By operating, the stream allocation update can be performed.

In Fig. 4, the flash translation layer 325 is shown as responsible for performing stream allocation and stream identifier update. However, in other embodiments according to the technical idea of the present invention, one or more of the components shown in FIG. 4 are implemented in software, for example, the memory controller 125 of FIG. 1 and the device driver of FIG. 1 ( 130), or implemented as library routines that can intercept write requests and assemble streams before issuing write commands, or Alternatively, it may be implemented as a separate special purpose hardware elsewhere in the machine 105. Although the description accompanying FIGS. 4-20 focuses on the implementation within the flash transformation layer 325, for the purposes of this discussion, reference to the stream allocation performed by the components of FIG. It is intended to contain an implementation of

In FIG. 4, the logical block address mapper 410 and the stream selection logic 415 are shown as separate components, but logically, the structures of these components may be combined into a single component. More generally, embodiments of the technical idea of the present invention combine certain components shown and described as being separate in FIGS. 4 to 6, 9 to 13, and 15 to 20 into a single component. can do.

FIG. 5 shows logical block addresses (LBAs) of various instructions mapped to chunk identifiers (IDs) and later mapped to stream identifiers for a solid state drive of FIG. 1. Embodiments of the inventive concept may support any number of write commands, but in FIG. 5, write commands 505-1, 505-2, and 505-3 are illustrated. In addition, although only write commands are shown in FIG. 5, embodiments of the inventive concept may also be applied to read commands. The write commands 505-1 to 505-3 may include logical block addresses 510-1 to 510-3, which are the starting logical block addresses of the write commands 505-1 to 505-3. Can be specified.

The logical block address mapper 410 may access the logical block addresses 510-1 to 510-3 from the write commands 505-1 to 505-3. When the logical block addresses 510-1 to 510-3 are read from the write commands 505-1 to 505-3, the logical block address mapper 410 is assigned to the corresponding chunk identifiers 515-1 to 515 -3) can be judged. Chunks may be considered as logical subdivisions of solid state drive 120 (but it is not necessary to arrange the chunks into other logical subdivisions of solid state drive 120, such as blocks or pages). The number of chunks is too much to be done in any desired way, for example by selecting a chunk of the desired size and dividing the size of the solid state drive 120 by the chunk size, or by the background logic 435 of FIG. 4. It can be determined by choosing a number that has proven to provide adequate usefulness that does not require much processing (and by dividing the solid state drive 120 into multiple chunks). Although the uniformity of size can simplify the operations of the logical block address mapper 410, the chunks need not have the same size. Thus, for example, one chunk may contain 128 KB of data in the solid state drive 120, while another chunk may contain 512 KB of data in the solid state drive 120. Since determining a chunk from a logical block address is performed using shift operations, chunk sizes that are powers of two are particularly advantageous. However, embodiments of the inventive concept may support chunks of any required size.

The logical block address mapper 410 may determine the chunk identifiers 515-1 to 515-3 in a number of ways. For example, the logical block address mapper 410 includes an arithmetic logic unit (ALU) 520 and the chunk identifiers 515-1 from the logical block addresses 510-1 to 510-3. ~515-3) can be calculated mathematically. Alternatively, the logical block address mapper 410 extracts various bits from the logical block addresses 510-1 to 510-3 to provide an address mask 525 that leaves the chunk identifiers 515-1 to 515-3. Can include.

When the chunk identifiers 515-1 to 515-3 are determined, the stream selection logic 415 may determine a stream to be used in the corresponding chunk. The stream selection logic 415 may determine corresponding stream identifiers 530-1 to 530-3 from the chunk-stream mapper 340 using the chunk identifiers 515-1 to 515-3. .

6 is the commands 505-1 to 505-2 of FIG. 5 modified to include the stream identifiers 530-1 and 530-2 of FIG. 5 and transmitted to the solid state drive 120 of FIG. 1 (For space reasons, FIG. 6 shows only two of the three write commands from FIG. 5 and associated data. However, FIG. 6 is generalized in the same manner as in FIG. 5). In FIG. 6, the stream identifier adder 420 may write the stream identifiers 530-1 and 530-2 to the write commands 505-1 and 505-2. In this way, the solid state drive 120 of FIG. 1 can know which device stream to use when processing the write commands 505-1 and 505-2. The transmitter 425 may transmit the modified write commands 505-1 and 505-2 to the solid state drive 120 of FIG. 1.

Not shown in FIGS. 5 and 6 is the operation of the line erecting device 430 of FIG. 4. When the chunk identifiers 515-1 to 515-3 of FIG. 5 are determined, the line up 430 of FIG. 4 is performed by the chunk identifiers of FIG. 5 for later processing of the background logic 435 of FIG. 515-1 to 515-3) may be lined up in the submission queue 335 of FIG. 3. The line up 430 of FIG. 4 may line up the chunk identifiers 515-1 to 515-3 of FIG. 5 at any time after the chunk identifiers 515-1 to 515-3 of FIG. 5 are determined. . For example, after the chunk identifiers 515-1 to 515-3 of FIG. 5 are determined by the logical block address mapper 410 of FIG. Chunk identifiers 515-1 to 515 of FIG. 5 after instructions 505-1 to 505-3 have been transmitted by transmitter 425 to solid state drive 120 of FIG. 1, or at any time in between. -3) can be lined up.

FIG. 7 shows the arithmetic logic unit (ALU) 520 of FIG. 5 that maps the logical block address 510-1 of FIG. 5 to the chunk identifier 515-1 of FIG. 5. In FIG. 7, the arithmetic logic unit 520 may receive a logical block address 510-1 and a chunk size 705. By dividing the logical block address 510-1 by the chunk size 705 (and discarding any fractional part of the calculation), the chunk identifier 515-1 can be calculated.

As described above with reference to FIG. 4, the background logic 435 of FIG. 4 can be implemented in a variety of ways. 8 to 13 illustrate an embodiment of the technical idea of the present invention, and FIGS. 14 to 20 describe another embodiment of the technical idea of the present invention. These embodiments of the technical idea of the present invention are not intended to be mutually exclusive, and a combination of the two embodiments of the technical idea of the present invention is possible. In addition, whether explicitly described herein or not, other embodiments of the technical idea of the present invention are also possible.

In the first embodiment of the technical idea of the present invention for the background logic 435 of FIG. 4, the chunk-stream mapper 340 of FIG. 3 is a sequential, frequency, and latest (Sequential, Frequency, Recency) (SFR) table. It may include. "Sequence, frequency, latest" refers to how the stream identifier to be assigned to the chunk is determined and updated. "Sequential" refers to a situation in which one write command is sequential with another write command. That is, the second write command writes data to the next logical block address after the previous write command. If the two write commands are sequential, it makes sense that both commands must be assigned to the same stream identifier.

As described below with reference to FIG. 10, write commands are issued from many different software sources, so the "sequence" in this context should not be read completely literally. Interrupting write commands when two "sequential" write commands from a single software source (eg, a database application) are separated by a write command from some other software source (eg, a file system). Should not rule out the sequence of two write commands from the database application. Indeed, a way to avoid preventing intervening write commands from interfering with the sequence is provided here.

"Frequency" refers to how often the chunk is accessed. The more frequently a chunk is accessed, the hotter the chunk is considered hotter (the temperature acts like a priority. The higher the temperature of the chunk, the higher the priority should be). Thus, the more frequently a chunk is accessed, the higher the stream priority should be assigned.

"Latest" refers to how old the chunk has been since it was last accessed. "Latest" acts as a balance for "frequency". The longer the period after the chunk was last accessed, the cooler the chunk is considered. Therefore, chunks should be assigned a lower priority. Thus, "frequency" increases the associated stream of a chunk, while "latest" decreases the associated stream of a chunk. "Sequential", "frequency", and "latest" are further described below with reference to FIGS. 10 to 13.

Moving to an embodiment of the technical idea of the present invention shown in FIG. 8, FIG. 8 shows the chunk identifiers 515-1 to 515-3 of FIG. 5 according to an embodiment of the technical idea of the present invention. It shows an exemplary SFR table that can be used to map to s 530-1 to 530-3. In FIG. 8, the SFR table 805 may store various data. For example, the SFR table 805 includes various chunk identifiers 515-1 to 515-4 and SFR corresponding stream identifiers 530-1 to 530- for each chunk identifier 515-1 to 515-4. 4) can be saved.

In addition to the stream identifiers 530-1 to 530-4, the SFR table 805 may store additional data. For example, for each chunk identifier 515-1 to 515-4, the SFR table 805 may store the current (ie, the most recent) access time 810-1 to 810-4 for the chunk. . The current access times 810-1-810-4 may be the time of the last access of a chunk of any type (eg, reads or writes) or only the most recent write command. The SFR table 805 may also store a previous access time 815-1 through 815-4 indicating a time when the chunk was previously accessed. Finally, the SFR table 805 may store access counts 820-1 to 820-4 indicating the number of accesses in the chunk. As described above, the access counts 820-1 to 820-4 may represent the total number of accesses (both reads and writes) of the chunk or only the total number of write accesses of the chunk.

Although FIG. 8 shows one possible embodiment of the technical idea of the present invention, other embodiments of the technical idea of the present invention may store more or less data than that shown in FIG. 8. Also, data structures other than tables can be used to store information. Embodiments of the technical idea of the present invention are intended to include all these changes.

9 shows additional details of the flash conversion layer 325 of FIG. 3 and the background logic 435 of FIG. 4 according to the first embodiment of the inventive concept. In FIG. 9, the flash transform layer 325 may also include sequential logic 905. The sequential logic 905 may determine whether the write command 505-1 of FIG. 5 is sequential with some previous write commands in order to use the same stream allocation as the previous write command. Sequential logic 905 is further discussed with reference to FIG. 10 below.

Background logic 435 may include state-of-the-art logic 910, access count adjuster 915, and stream identifier adjuster 920. The latest logic 910 may calculate the latest recency weight for the chunk. The access count adjuster 915 may adjust the access count for the stream based on the latest weighting. Further, the stream identifier adjuster 920 may calculate a new stream identifier to be used for the chunk based on the adjusted access count. The state-of-the-art logic 910, the access count adjuster 915, and the stream identifier adjuster 920 perform arithmetic calculations, so they can all be implemented using one (or more) separate or shared arithmetic logic units. Otherwise, state-of-the-art logic 910, access count adjuster 915, and stream identifier adjuster 920 can be implemented using special purpose hardware designed to compute only the specific functions described here and not compute anything else. have.

10 shows details of the sequential logic 905 of FIG. 9. In FIG. 10, sequential logic 905 is shown as including storage 1005 and comparator 1010. For practical goals, storage 1005 may be part of storage 330 of FIG. 3, rather than a separate storage within sequential logic 905. The storage 1005 may store a window 1015, which is information on a set of latest write commands, and each of the write commands may have a stream identifier assigned to it. For example, FIG. 10 shows entries 1010-1 to 1020-8, of which entries 1010-1 to 1020-4 exist in the window 1015. Entries 1010-1-1020-8 may be managed in the queue 1025. Queue 1025 may take any form required, for example, in the form of an array or linked list with a pointer to the front of the queue 1025, naming the two exemplary implementations. Each of the entries 1010-1 to 1020-8 may include an end logical block address 1030 and a stream identifier 1035 for a previous write command. Window 1015 may include a window size 1040 that specifies how many recent entries are included in window 1015. In FIG. 10, the window size 1040 is shown to include four entries, but embodiments of the inventive concept may support many entries in the window 1015. The window size 1040 is, for example, the number of cores of the processor of FIG. 1 or software sources (operating system, file system, applications, etc.) executed on the processor 110 of FIG. 1 that can issue write instructions. It can depend on several factors including the number of. Embodiments of the inventive concept may support window sizes determined based on other factors as well. Further, the window size 1040 may be set statically or may be dynamically changed as conditions in the machine 105 of FIG. 1 change.

When the new write command 505-1 of FIG. 5 is received, the comparator 1010 transfers the logical block address 510-1 of FIG. 5 to the end of the entries 1020-1 to 1020-4 of the window 1015. It can be compared with the logical block addresses 1030. If the logical block address 510-1 of FIG. 5 is the next address after the last logical block address 1030 of the previous write command, the logical block address 510-1 of FIG. 5 is the last logical block address 1030 of the previous write command. ) Can be considered sequential. Otherwise, if there is no valid logical block address that can be used between the last logical block address 1030 of the previous write command and the logical block address 510-1 of FIG. 5, the logical block address 510-1 of FIG. 5 is The end logical block address 1030 of the previous write command may be regarded as sequential (if there are no possible intervening addresses where data could be written, the end logical address 1030 is literally written by the previous write command. It should be noted that it does not have to be the last address). If the logical block address 510-1 of FIG. 5 is sequential to the end logical block address 1030 of any entry 1020-1-1020-4 of the window 1015, the entries 1010-1-1020-4 A stream identifier 1035 of) is used for the write command 505-1 of FIG. 5.

Identifying the sequential logical block addresses is one method of determining when the sequence has occurred, but embodiments of the inventive concept may support other implementations for detecting the sequence. For example, sequential logic 905 may use a pattern matcher to detect when logical block addresses are accessed for a particular software source in a repeating pattern.

Regardless of whether or not the logical block address 510-1 of FIG. 5 is sequential to the end logical block address 1030 of any of the entries 1010-1 to 1020-4 in the window 1015, The oldest entry is "ejected", and the write command 505-1 of FIG. 5 can be added to the window 1015. For example, it is assumed that the entry 1020-1 is the entry for the most recent write command in the window 1015 and the entry 1020-4 is the oldest entry in the window 1015. The entry 1020-4 is removed from the window 1015, and the logical block address 510-1 and the stream identifier 530-1 of FIG. 5 may be added to the window 1015 as new entries. The window 1015 can be implemented in any required way. For example, window 1015 may contain a circular list (such as an array) with pointers to the oldest entries. When a new entry is added to window 1015, the oldest entry specified by the pointer is overwritten by the new entry, and the pointer can be adjusted to point to the next oldest entry in window 1015. Embodiments of the inventive concept may support any necessary structure for storing information about the window 1015. The circular list is just one of these possible data structures.

FIG. 11 shows the calculation of the latest weighting using the latest logic of FIG. 9. Referring to FIG. 11, the latest logic 910 is shown to calculate the latest weight 1105. The formula shown in Figure 11 is the difference between the current access time 810-1 (i.e., the most recent) and the previous access time 815-1 for the chunk, divided by the decay period 1110 to the power of 2 Calculate the latest weight (1105) as The decay period 1110 represents an adjustable variable that controls how quickly a chunk is degraded to lower priority streams. The decline period 1110 is assigned an initial value when the machine 105 of FIG. 1 starts, and can be adjusted as desired (manually by the system administrator or automatically depending on the system workload). Preventing chunks from being promoted too quickly (which causes most of the chunks to use the same high priority stream) or being promoted too slowly (which causes most of the chunks to use the same low priority stream). It is desirable. In other words, it is desirable to have the chunks allocated to the streams in a very uniform way. No individual stream should be used too hard or too lightly. The decline period 1110 represents how to manage stream promotion and demotion to achieve this goal.

It should be noted that the latest weight 1105 may be different for each chunk. On the other hand, the decay period 1110 should be constant across all calculations for the latest weight 1105 (but this does not mean that the decay period 1110 does not change over time).

FIG. 12 shows that the access count adjuster 915 of FIG. 9 adjusts the access count using the latest weighting 1105 of FIG. 11. In FIG. 12, the access count adjuster 915 is shown calculating the adjusted access count 1205. The adjusted access count 1205 may be calculated as one more than the access count 820-1 divided by the latest weight 1105. The adjusted access count 1205 may be stored back in the location of the access count 820-1, for example in the SFR table 805 of FIG. 8.

FIG. 13 shows that the stream identifier controller 920 of FIG. 9 is used to calculate the stream identifier from the adjusted access count 1205 of FIG. 12. In FIG. 13, the stream identifier adjuster 920 is shown to calculate the stream identifier 530-1. The stream identifier 530-1 is calculated as a log of the adjusted access count 1205, and may be stored again in the location of the stream identifier 530-1, for example, the SFR table 805 of FIG. have. The mathematical term “log” when used for computers typically means log ₂ or log ₁₀ , but a log function using any desired base may be chosen. The radix selected for the logarithmic function is that embodiments of the technical idea of the present invention avoid chunks being promoted too quickly or too slowly (and thus excessive use of one or more device streams and poor use of other device streams). It provides another mechanism. Stream identifier adjuster 920 may also use more than one log function depending on how large an adjusted access count 1205 is obtained. For example, if the adjusted access count 1205 is lower than some threshold, the stream identifier adjuster 920 may calculate the stream identifier 530-1 using _{log 2.} If the adjusted access count 1205 is greater than the corresponding threshold, the stream identifier adjuster 920 may calculate the stream identifier 530-1 using _{log 10.}

It is worth making a couple of references to FIGS. 11 to 13. First, FIGS. 11 to 13 show specific calculations for calculating the latest weight 1105 of FIG. 11, the adjusted access count 1205 of FIG. 12, and the stream identifier 530-1. However, embodiments of the inventive concept may support calculating them (or other values) in any desired method. Since the ultimate goal is to be able to properly adjust the stream identifier 530-1 of Fig. 8 for the chunk 515-1 of Fig. 8, any desired attempt to achieve this result can be used. Figures 11-13 only show one such attempt.

Second, in FIGS. 11 to 13, the access times (such as the current access time 810-1 and the previous access time 815-1 in FIG. 11) are determined based on the number of issued requests rather than a specific clock. Can be. For example, it is assumed that the write commands 505-1 to 505-3 of FIG. 5 are issued continuously. The write command 505-1 of FIG. 5 may be a fourth write command, the write command 505-2 of FIG. 5 may be a fifth write command, and the write command 505-3 of FIG. 5 is It could be the sixth write command. In this example, the "access times" for the various write commands may be "4", "5", and "6", respectively.

This choice of judging access times means that if all software sources stop sending write commands for some time interval (such as 1 second, 1 minute, or 1 hour) of time, the chunks are not technically accessed for a significant amount of time. The "temperature" of the chunks has an unchanging result. Thus, even if a chunk has not been accessed for a significant amount of time, the "temperature" of that chunk does not change if another chunk is not accessed for that significant time. From a practical point of view, it is unlikely that all software sources will stop sending write commands at the same time. However, when it occurs, embodiments of the technical idea of the present invention can handle this situation. In addition, embodiments of the inventive concept may support the use of clock time instead of instruction numbers, which results in cooling of chunks even if software sources do not issue write instructions.

Contrary to FIGS. 8 to 13, FIGS. 14 to 20 show another embodiment of the technical idea of the present invention. Like embodiments of the inventive concept illustrated in FIGS. 8 to 13, FIGS. 14 to 20 illustrate the chunk identifiers 515-1 to 515-3 of FIG. 5 and the stream identifiers 530-of FIG. 5. 1 to 530-3) and the background logic 435 of FIG. 4 may be implemented.

14 illustrates mapping of chunk identifiers 515-1 to 515-3 of FIG. 5 to stream identifiers 530-1 to 530-3 of FIG. 5 according to a second embodiment of the technical idea of the present invention. Shows the node used to do it. In FIG. 14, nodes 1405-1 to 1405-3 are shown, and node 1405-1 is shown in detail. Nodes 1405-1 to 1405-3 can be stored in any desired manner, such as arrays, linked lists, and other data structures. In the solid state drive 120 of FIG. 1, there may be one node for each chunk.

The node 1405-1 may include various data including a chunk identifier 515-1, a stream identifier 530-1, an access count 820-1, and an expiration time 1410. The chunk identifier 515-1, the stream identifier 530-1, and the access count 820-1 store data similar to those elements of the first embodiment of the inventive concept. The expiration time 1410 represents the time at which the chunk will be considered expired due to lack of accesses by write commands. As in the first embodiment of the inventive concept, "time" may be measured as the number of specific write commands in the order of all write commands rather than measuring the clock.

15 shows details of the background logic 435 of FIG. 4 according to a second embodiment of the inventive concept. In FIG. 15, background logic 435 includes promotion logic 1505, second line up 1510, storage 1515 for queues 1521-1520-3, and down logic 1525. Includes. Promotion logic 1505 may promote the chunk to a higher priority stream when appropriate. The second line up device 1510 (named to distinguish it from the line up device 430 of FIG. 4, even though the operation is similar) queues the chunk identifiers 515-1 to 515-3 of FIG. 5 ( 1520-1~1520-3). The second line up device 1510 may have a structure similar to the line up device 430 of FIG. 4. Like storage 1005 of FIG. 10, storage 1515 may be substantially part of storage 330 of FIG. 3 rather than a separate storage within background logic 435. Embodiments of the inventive concept may support any number of queues. The three queues 1520-1 to 1520-3 are only examples. In addition, one queue may exist for each of the stream identifiers 530-1 to 530-3 of FIG. 5. That is, it is a coincidence that the number of queues 150-1 to 1520-3 and the number of chunk identifiers 515-1 to 515-3 of FIG. 3 are the same. When the chunk identifiers 515-1 to 515-3 of FIG. 5 reach the front of the queues 150-1 to 1520-3, the degrading logic 1525 is performed by the chunk identifiers 515-1 to 5 of FIG. 51503 (It can be judged whether to downgrade.

16 shows promotion and demotion of chunk identifiers in the queues 1520-1 to 1520-3 of FIG. 15. In FIG. 16, the chunk identifiers 515-1 to 515-3 of FIG. 3 are expressed using characters a to d. Unnamed entries of the queues 1520-1 to 1520-3 are not used in the example shown in FIG. 16, and if not used separately in this example, an arbitrary chunk identifier may be stored.

First, in FIG. 16, a submission queue 335 with a chunk identifier (a) at the forefront 1605 of the submission queue 335 is shown. The chunk identifier (a) may identify a chunk that has not been accessed previously. Thus, after the chunk identifier (a) is removed from the submission queue (335), as shown by the dashed arrow (1610), the chunk identifier (a) is the tail of the queue (1520-1) for the "coolest" stream. Can be added to.

Secondly, in FIG. 16, the submission queue 335 may include a chunk identifier b (which may be moved to the front of the submission queue 335 after the chunk identifier a is processed (1605)). . As a result of increasing the access count 820-1 of FIG. 14 with respect to the chunk identifier (b), the chunk identifier (b) can be promoted to a "hotter" stream. As shown by the dotted arrow 1615, after the chunk identifier b is removed from the submission queue 335, the chunk identifier b is transferred from the entry 1620 of the queue 1520-1 to the queue 1520-2 ) Can be moved to the tail.

Third, the downgrade logic 1525 of FIG. 5 may alternately check the fronts 1625-1 to 1625-3 of the queues 150-1 to 1520-3. For example, the downgrade logic 1525 of FIG. 15 may check the chunk identifier c of the front 1625-3 of the queue 1520-3. If the chunk identifier c is going to be degraded, the chunk identifier c is moved from the front of the queue 1520-3 to the tail of the queue 1520-2 (the queue for the next "hottest" stream). As a result of the chunk identifier (c) being degraded, as described below with reference to FIG. 17, the chunk identifier (d) of the queue 1520-3 may be the hottest chunk.

17 shows details of the promotion logic 1505 of FIG. 15. In FIG. 17, the promotion logic 1505 may include an increaser 1705, a stream identifier adjuster 1710, a chunk expiration logic 1715, and a hottest chunk logic 1720. The increaser 1705 may increase the access count 820-1 of FIG. 14 to produce an adjusted access count. The stream identifier adjuster 1710 may calculate the stream identifier 530-1 of FIG. 5 from the adjusted access count as illustrated in FIG. 18. The stream identifier adjuster 1710 may be structurally similar to the stream identifier adjuster 920 of FIG. 9.

Returning to FIG. 17, the chunk expiration logic 1715 may determine an expiration time for the chunk. The expiration time for the chunk may be determined as the current access time for the chunk identifier plus the device lifetime. Device lifetime is an interval of time depending on which chunk is the hottest chunk, discussed below. 19 shows this as an equation. The expiration time 1410 is the sum of the access count 820-1 (the current access time for the chunk identifier 515-1 of FIG. 5, as discussed above) and the device lifetime 1905.

Finally, referring again to FIG. 17, the hottest chunk logic 1720 may determine whether the chunk identifier 515-1 of FIG. 5 represents the currently hottest chunk. The hottest chunk may be defined as a chunk identifier having the highest stream priority. If there are multiple chunk identifiers having the highest stream priority, a random selection algorithm is used so that a specific chunk identifier having the hottest stream priority can be selected as the hottest chunk. For example, the first chunk identifier of the queue indicating the hottest stream priority may be selected as the hottest chunk. Alternatively, the chunk identifier having the most recent access (ie, the highest current access time) may be selected as the hottest chunk. Alternatively, the chunk identifier of the tail of the queue indicating the hottest stream priority may be selected as the hottest chunk. If the chunk identifier 515-1 of FIG. 5 is the hottest chunk, the device lifetime 1905 of FIG. 19 is the current access time of the chunk identifier 515-1 of FIG. 5 and the chunk identifier 515-1 of FIG. 5 Is calculated as the difference between the previous access times. The device lifetime 1905 of FIG. 19 may represent a predicted maximum interval between two write commands of a specific chunk, and may be defined based on the interval between the two most recent write commands for the hottest chunk. The device lifetime 1905 of FIG. 19 may be used to determine the expiration time 1410 of FIG. 14 that may have an effect when the chunk identifier 515-1 of FIG. 14 is degraded.

20 shows details of the downgrade logic 1525 of FIG. 15. In FIG. 20, the degrading logic 1525 is shown as including a comparator 2005 and a reducer 2010. The comparator 2005 may compare the current time (again, which may be the number of the most recent write requests rather than the clock time) with the expiration time 1410 of FIG. 14. If the current time has passed the expiration time 1410 of FIG. 14, the chunk may be considered cooler than before because it has been left for a time interval that can be measured by the device lifetime 1905 of FIG. 19 without write commands. In other words, if the device lifetime 1905 of FIG. 19 represents the number of predicted write instructions between writes for any given chunk, then at the most recent access time of the chunk identifier 515-1 of FIG. The expiration time 1410 of FIG. 14, calculated as adding the device lifetime 1905 of FIG. 14, may represent the last time that the chunk identifier 515-1 of FIG. 14 was written and is not considered cooled. After the expiration time 1410 of FIG. 14 has elapsed, the chunk identifier 515-1 of FIG. 14 has cooled down somewhat and may be degraded. In this case, the reducer 2010 reduces the stream identifier 530-1 of FIG. 5 to correspond to the reduced "temperature" of the chunk identifier 515-1 of FIG. 5 to reduce the chunk identifier 515-1 of FIG. ) Priority can be reduced.

If the expiration time 1410 of FIG. 14 has not passed, the chunk identifier 515-1 of FIG. 14 has not yet been cooled, and the chunk identifier 515-1 of FIG. 14 may be left at the front of the queue. It should be noted that this does not mean that other chunk identifiers in the queue will not be degraded. Since the chunk identifiers are placed in the queues in the order in which the chunks are accessed, the chunk identifier 515-1 of FIG. 14 is the chunk identifier that was accessed the oldest (among the chunk identifiers of the queue), and thus the first chunk to expire. It will be an identifier. When the chunk identifier 515-1 of FIG. 14 is accessed again (and thus remains "hot"), the chunk identifier 515-1 of FIG. 14 is moved to the tail of the queue, and another chunk identifier is at the front of the queue. It is left behind and is therefore potentially degraded.

21A and 21B are flowcharts of exemplary procedures for determining the stream identifier 530-1 of FIG. 5 with respect to the write command 505-1 of FIG. 5 according to an embodiment of the inventive concept. . In FIG. 21A, at block 2105, the receiver 405 of FIG. 4 may receive a write command 505-1 from a software source. At block 2110, the logical block address 510-1 of FIG. 5 may be read from the write command 505-1 of FIG. 5. In block 2115, the logical block address mapper 410 of FIG. 4 may map the logical block address 510-1 of FIG. 5 to the chunk identifier 515-1 of FIG. 5. As described above, the logical block address mapper 410 of FIG. 5 can be configured in any desired manner, for example, by filtering specific bits from the logical block address 510-1 of FIG. The chunk identifier 515-1 of FIG. 5 may be determined by dividing the address 510-1 by the chunk size 705 of FIG. 7. At block 2120, the stream selection logic 415 of FIG. 5 may select the stream identifier 530-1 to be used for the chunk identifier 515-1 of FIG. 5. As described above, the stream selection logic 415 of FIG. 5 may be performed in any necessary manner, e.g., by retrieving the stream identifier 530-1 of FIG. 5 from the chunk-stream mapper 340 of FIG. , By calculating the stream identifier 530-1 of FIG. 5 from the access count 820-1 of FIG. 8, allocating device streams in a round robin manner, or in any other necessary attempt. In block 2125, the stream identifier adder 420 of FIG. 4 may add the stream identifier 530-1 of FIG. 5 to the write command 505-1 of FIG. 5. Finally, at block 2130, the write command 505-1 of FIG. 5 is the transmitter 425 of FIG. 4 that transmits the write command 505-1 of FIG. 5 to the solid state drive 120 of FIG. It is processed by the solid state drive 120 of FIG. 1, which may include, to perform a designated write operation.

At this time, the write command 505-1 of FIG. 5 is completely processed. However, other processing such as updating the stream identifier 530-1 of FIG. 5 assigned to the chunk identifier 515-1 of FIG. 5 is still performed. At block 2135 (FIG. 21A), the line up 430 of FIG. 4 may add the chunk identifier 515-1 of FIG. 5 to the submission queue 335 of FIG. 3 for further processing. At block 2140, when the chunk identifier 515-1 of FIG. 5 reaches the front of the submit queue 335 of FIG. 3, the chunk identifier 515-1 of FIG. ) Can be removed from. Finally, at block 2145, the background logic 435 of FIG. 4 may update the stream identifier 530-1 of FIG. 5 with respect to the chunk identifier 515-1 of FIG. 5. Various attempts for the background logic 435 of FIG. 4 to update the stream identifier 530-1 of FIG. 5 to the chunk identifier 515-1 of FIG. 5 are described below with reference to FIGS. 24 and 25A-25C. It is described in.

22 is a logical block address mapper 410 of FIG. 4 that maps the logical block address 510-1 of FIG. 5 to the chunk identifier 515-1 of FIG. 5 according to an embodiment of the inventive concept. Shows a flow chart of an exemplary procedure for. In FIG. 22, at block 2205, the logical block address mapper 410 of FIG. 4 filters bits from the logical block address 510-1 of FIG. 1 using the address mask 525 of FIG. The chunk identifier 515-1 of 5 may be left. Otherwise, at block 2210, the logical block address mapper 410 of FIG. 4 divides the logical block address 510-1 of FIG. 5 by the chunk size 705 of FIG. 7 and the chunk identifier 515 of FIG. -1) can be judged.

23A and 23B are for determining the stream identifier 530-1 of FIG. 5 with respect to the chunk identifier 515-1 of FIG. 5 using sequential logic according to the first embodiment of the technical idea of the present invention. A flowchart of an exemplary procedure is shown. In FIG. 23A, at block 2305, the stream selection logic 415 of FIG. 4 may determine whether the write commands 505-1 to 505-3 are being tested for sequential logical block addresses. For example, in embodiments of the inventive concept using a multi-queue attempt, the stream selection logic 415 of FIG. 4 considers whether the write command 505-1 of FIG. 5 is sequential to the previous write commands. I won't. If the stream selection logic 415 of FIG. 4 is not testing for sequential write instructions, then at block 2315, the stream selection logic 415 of FIG. By accessing the stream identifier 530-1 of FIG. 5 from 340, it is possible to determine the stream identifier 530-1 of FIG. 5 that is currently associated with the chunk identifier 515-1 of FIG. 5. Details of how the stream selection logic 415 of FIG. 4 determines the stream identifier 530-1 of FIG. 5 may depend on how the chunk-stream mapper 340 of FIG. 3 is implemented. If the chunk-stream mapper 340 includes the SFR table 805 of FIG. 8, the stream selection logic 415 of FIG. 4 may perform a table search. When the chunk-stream mapper 340 includes the nodes 1405-1 to 1405-3 of FIG. 14, the stream selection logic 415 of FIG. 4 determines the stream identifier 530-1 of FIG. Before, it will be necessary to search for nodes to find the chunk identifier 515-1 of FIG. 5.

On the other hand, if the stream selection logic 415 of FIG. 4 is testing for sequential write commands, at block 2315, the stream selection logic 415 of FIG. The window 1015 of FIG. 10 of 8) can be identified. At block 2320, the stream selection logic 415 of FIG. 4 indicates that the logical block address 510-1 of FIG. 5 is assigned to any of the entries 1021-1020-8 of FIG. 10 in window 1015 of FIG. It may be determined whether it is sequential to the last logical block address 1030 of FIG. 10 for ). The logical block address 515-1 of FIG. 5 is sequential to the last logical block address 1030 of FIG. 10 for any of the entries 1020-1 to 1020-8 of FIG. 10 in the window 1015 of FIG. Otherwise, in block 2310, the stream selection logic 415 of FIG. 4 may determine the stream identifier 530-1 of FIG. 5 currently associated with the chunk identifier 515-1 of FIG. 5 as described above. .

The stream selection logic 415 of FIG. 4 is being tested for sequential write commands, and the logical block address 510-1 of FIG. 5 is an arbitrary entry 1010-1-of FIG. 10 in the window 1015 of FIG. 1020-8), in block 2325 (FIG. 23B), the stream selection logic 415 of FIG. 4 may determine the stream identifier 530-1 of FIG. 5 assigned to the previous write command. At block 2125 (which is part of FIGS. 21A and 21B, and thus shown by dotted lines in FIG. 23B for explanatory purposes), the stream identifier 530-1 of FIG. 5 assigned to the previous write command is shown in FIG. May be assigned to the write command 505-1 of. At block 2330, the stream selection logic 415 of FIG. 4 may identify entries 1010-1-1020-8 of FIG. 10, which are the oldest entries in window 1015 of FIG. 10. Finally, at block 2335, the stream selection logic 415 of FIG. 4 removes the oldest entry in the window 1015 of FIG. 5 and creates a new entry corresponding to the write command 505-1 of FIG. Can be added. The technique of how the stream selection logic 415 of FIG. 4 performs this deletion and addition depends on the structure used in the window 1015 of FIG. 10. If the window 1015 of FIG. 10 stores the array or linked list of the entries 1010-1 to 1020-8 of FIG. 10, the deletion and addition overwrite the oldest entry with a new value, and the array or concatenation of the pointer. This can include things like updating to the beginning of the list. On the other hand, if a different structure is used for the window 1015 of FIG. 10, the deletion and addition delete and deallocate the memory object for the oldest entry, and the new memory object for the write command 505-1 of FIG. May include allocating.

24 is a flowchart of an exemplary procedure for performing a background update of the SFR table 805 of FIG. 8 according to the first embodiment of the inventive concept. In FIG. 24, at block 2405, the access count adjuster 915 of FIG. 9 may increment the access count 820-1 of FIG. 8 (eg, using an incrementer of arithmetic logic). At block 2410, the latest logic 910 of FIG. 9 may calculate the latest weight 1105 of FIG. 11 (eg, using arithmetic logic). At block 2415, the access count adjuster 915 of FIG. 9 is the access count 820-1 of FIG. 8 (incremented at block 2405) of FIG. 11 (e.g., using arithmetic logic). It can be divided by the latest weight (1105). Finally, at block 2420, the stream identifier adjuster 920 of FIG. 9 may determine a new stream identifier to be associated with the chunk identifier 515-1 of FIG. 5. For example, the stream identifier adjuster 920 of FIG. 9 may calculate the log of the adjusted access count 1205 of FIG. 12 using an arithmetic logic unit.

24A to 25C are flowcharts of exemplary procedures for performing background update of node 1405-1 of FIG. 14 according to a second embodiment of the inventive concept. In FIG. 25A, at block 2505, the increaser 1705 of FIG. 17 may increment the access count 820-1 of FIG. 14. At block 2510, the chunk expiration logic 1715 of FIG. 17 may determine the expiration time 1410 of FIG. 14 for the chunk identifier 515-1 of FIG. 14. At block 2515, the stream identifier adjuster 1710 of FIG. 17 may determine the stream identifier 530-1 of FIG. 14 with respect to the chunk identifier 515-1 of FIG. 14. The stream identifier controller 1710 of FIG. 17 may determine the stream identifier 530-1 of FIG. 14 by accessing the stream identifier 530-1 of FIG. 14 from the node 1405-1 of FIG. 14, for example. I can. In block 2520, the second line up 1510 of FIG. 15 is assigned to one of the queues 1520-1 to 1520-3 of FIG. 15 corresponding to the stream identifier 530-1 of FIG. 14. The chunk identifier 515-1 of may be placed.

In block 2525 (FIG. 25B), the hottest chunk logic 1720 of FIG. 17 uses the access count 820-1 of FIG. 14 for the chunk identifier 515-1 of FIG. 14 to access the hottest chunk. It can be compared to the count. If the access count 820-1 of FIG. 14 for the chunk identifier 515-1 of FIG. 14 is greater than the access count for the hottest chunk, the chunk identifier 515-1 of FIG. 14 is now the hottest chunk. Thus, at block 2530, the hottest chunk logic 1720 of FIG. 17 reveals that the chunk identifier 515-1 of FIG. 14 is the new hottest chunk, and at block 2535, the hottest chunk logic of FIG. 17 1720 may determine the device lifetime 1905 of FIG. 19 as a time difference between the two most recent accesses of the chunk identifier 515-1 of FIG. 14.

Regardless of whether the chunk identifier 515-1 of FIG. 14 is the hottest chunk, at block 2540, the comparator 2005 of FIG. 20 indicates that the chunk identifier 515-1 of FIG. 1520-1 to 1520-3), it may be determined whether the expiration time 1410 of FIG. 14 has passed. Otherwise, background logic 435 of FIG. 4 waits until expiration time 1410 of FIG. 14 for chunk identifier 515-1 of FIG. 14 has passed (doing other things in the meantime) and at block 2540. You can check again. At this time, in block 2545, the degrading logic 1525 of FIG. 15 may remove the chunk identifier 515-1 of FIG. 14 from the queues 150-1 to 1520-3 of FIG. 15. Thereafter, at block 2550, the reducer 2010 of FIG. 20 may reduce the stream identifier 530-1 of FIG. 14.

In block 2555 (FIG. 25C), the second line up 1510 of FIG. 15 replaces the chunk identifier 515-1 of FIG. 14 with the queues 1521-1 to 1520 of FIG. 15 corresponding to the new stream identifier. It can be placed in the other of -3). At block 2560, the downgrade logic 1525 of FIG. 15 may determine whether the chunk identifier 515-1 of FIG. 14 was the hottest chunk. If so, at block 2565, the downgrade logic 1525 of FIG. 15 may select the other chunk identifier as the new hottest chunk. For example, another chunk assigned to stream identifier 530-1 of FIG. 14, such as a chunk identifier at the front or end of the queue, may be selected (before decrementing at block 2560). The downgrade logic 1525 of FIG. 15 may also, as before, calculate the device lifetime 1905 of FIG. 19 based on the selected new hottest chunk.

21A to 25C, some embodiments of the technical idea of the present invention are shown. However, those skilled in the art will recognize that other embodiments of the technical idea of the present invention are also possible by changing the order of blocks, omitting blocks, or including connections not shown in the drawings. All of these changes in the flow charts, whether explicitly described or not, are considered to be embodiments of the technical idea of the present invention.

The following description is intended to provide a concise and general description of a suitable machine or machines in which the technical idea of the present invention is implemented. Machines or machines are not only input from conventional input devices such as a keyboard, mouse, etc., but also instructions received from another machine, interaction with a virtual reality (VR) environment, biofeedback, or other input signals. At least some can be controlled. As used herein, the term “machine” may broadly encompass a single machine, a virtual machine, or a system of communicatively coupled machines, virtual machines, or devices operating together. Exemplary machines include computing devices such as personal computers, workstations, servers, portable computers, handheld devices, telephones, tablets, etc., as well as transport devices such as personal or public transport (e.g., cars, trains, taxis, etc.) Can include.

The machine or machines may include programmable or non-programmable logic devices or arrays, application specific integrated circuits (ASICs), embedded controllers such as embedded computers, smart cards, and the like. The machine or machines may utilize one or more connections to one or more remote machines, such as a network interface, modem, or other communication connection. Machines can be connected to each other through physical or logical networks such as intranet, Internet, local area network (LAN), wide area network (WAN), and so on. For those skilled in the field, network communications utilize a variety of wired or wireless short-range or long-range carriers and protocols, including radio frequency (RF), satellite, microwave, IEEE 802.11, Bluetooth®, optical, infrared, cable, laser, etc. It will be understood.

Embodiments of the present invention are associated with functions, procedures, data structures, application programs, etc. that allow the machine to perform tasks or define abstract data types or low-level hardware contexts when accessed by the machine. It can be described with reference to data. Associated data may be stored in volatile or nonvolatile memory such as RAM, ROM, etc. or other storage devices including hard drives, floppy disks, optical storage, tapes, flash memory, memory sticks, digital video disks, biometric storage, etc., and associated storage media. Can be saved. The associated data may be transmitted in the form of packets, serial data, parallel data, transmission signals, etc. over a transmission environment including a physical or logical network, and may be used in a compressed or encrypted form. Associated data can be used in a distributed environment and stored locally or remotely for machine access.

Embodiments of the present invention may include a real, non-transitory, machine-readable medium containing instructions that can be executed by one or more processors. Instructions may include instructions that cause elements of the invention to be performed as described herein.

Although the technical idea of the present invention has been described with reference to the embodiments, the illustrated embodiments can be modified and combined in a desired manner without departing from the technical idea of the present invention in terms of arrangement and detail. Although the foregoing descriptions have focused on specific embodiments, other configurations may also be considered. In particular, although the expression "according to an embodiment of the present invention" or a similar expression is used, these phrases refer to possible general embodiments, and are not intended to limit the technical idea of the present invention to the configuration of a specific embodiment. As used herein, these terms may refer to the same or different embodiments, which may be combined into the same or different embodiments.

The above-described embodiments do not limit the technical idea of the present invention. Although some embodiments have been described, those skilled in the art will understand that these embodiments may be variously modified without departing from the novel technical spirit and advantages of the present invention. Accordingly, all such modifications are included in the technical spirit scope of the present invention as defined in the claims.

Embodiments of the present invention may be extended to the following phrases, but are not limited thereto.

As a first phrase, an embodiment of the technical idea of the present invention includes a solid state drive (SSD). The solid state drive may include a flash memory for storing data; Support for a plurality of device streams in the solid state drive; A solid state controller including storage for a submission queue and a chunk-stream mapper, and managing writing data to the flash memory in response to a plurality of write commands; And it includes a flash conversion layer. The flash translation layer includes: a receiver for receiving a write command including a logical block address (LBA); A logical block address mapper for mapping the logical block address to a chunk identifier (ID); Stream selection logic for selecting a stream identifier based on the chunk identifier using the chunk-stream mapper; A stream identifier adder for adding the stream identifier to the write command; A line-up device for placing the chunk identifier in the submission queue; And background logic to lift the chunk identifier from the submission queue and update the chunk-stream mapper.

As a second phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the first phrase, and further includes a transmitter that transmits the write command to the solid state drive together with the stream identifier.

As a third phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the first phrase, and the logical block address mapper includes an address mask for determining the chunk identifier by covering a part of the logical block address. do.

As a fourth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the first phrase, and the logical block address mapper divides the logical block address by a chunk size and identifies the chunk identifier. (ALU, arithmetic logic unit) is included.

As a fifth phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the first phrase, and the chunk-stream mapper includes a sequential, frequency, and latest (SFR, Sequential, Frequency, Recency) table, , The SFR table includes the chunk identifier and the stream identifier for the chunk identifier.

As a sixth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the fifth phrase, and the background logic returns a previous stream if the logical block address is sequential to the second logical block address of the previous write command. Includes sequential logic to select.

As a seventh phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the sixth phrase, and the previous write command is in a window preceding the write command having a window size.

As an eighth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the seventh phrase, and the window size is the number of cores of the processor of the host computer system and the number of cores executed by the processor of the host computer system. It is determined in response to at least one of the number of software sources.

As a ninth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the seventh phrase, and the solid state drive controller further includes a storage for a queue of previous write commands including the previous write command. And, the queue includes an end logical address and a corresponding stream identifier for each of the previous write commands.

As a tenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the fifth phrase, and the background logic includes a current access time for the chunk identifier, a previous access time for the chunk identifier, and a decline. The latest logic to calculate the latest weight based on the period; An access count adjuster for producing an adjusted access count by adjusting an access count for the chunk identifier based on the latest weight; And a stream identifier adjuster for adjusting the stream identifier based on the adjusted access count for the chunk identifier.

As an eleventh phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the tenth phrase, and the SFR table includes the current access time for the chunk identifier, the previous access time for the chunk identifier, And further includes the access count for the chunk identifier.

As a twelfth phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the tenth phrase, and the latest logic determines the latest weight of 2 (the current access time for the chunk identifier and the chunk identifier). Dividing the difference between the previous access time with respect to the decay period) to the power.

As a thirteenth phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the tenth phrase, and the access count adjuster calculates the adjusted access count for the chunk identifier (the access count for the chunk identifier). Count plus 1) divided by the latest weight above.

As a fourteenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the tenth phrase, and the stream identifier controller calculates the stream identifier as a log of the adjusted access count for the chunk identifier. .

As a fifteenth phrase, an embodiment of the technical idea of the present invention includes a solid state drive according to the first phrase, the chunk-stream mapper includes a node entry, and the node entry includes the chunk identifier and the chunk identifier. Includes the stream identifier.

As a sixteenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the fifteenth phrase, and the background logic includes: a promotion logic for determining when to promote the stream identifier based on the chunk identifier; A second stringer for disposing the chunk identifier at the first of a plurality of queues corresponding to a plurality of stream identifiers in response to the stream identifier for the chunk identifier; And a degrading logic for determining when to degrade the stream identifier based on the chunk identifier.

As a seventeenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the sixteenth phrase, and the promotion logic includes: an increaser for increasing an access count for the chunk identifier; And a stream identifier adjuster for determining the stream identifier in response to the access count for the chunk identifier.

As the eighteenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the seventeenth phrase, and the stream identifier adjuster operates to determine the stream identifier based on a log of the access count for the stream identifier. .

As a nineteenth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the seventeenth phrase, and the promotion logic further includes a chunk expiration logic for calculating an expiration time for the chunk identifier.

As the twentieth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the nineteenth phrase, and the chunk expiration logic determines the expiration time for the chunk identifier and the access count for the chunk identifier and the device. Operates to calculate as the sum of the lifetimes.

As a 21st phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the 20th phrase, and the device lifetime is between a last access time for the hottest chunk and a previous access time for the hottest chunk. It's the difference.

As the 22nd phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the 21st phrase, and in the promotion logic, the access count for the chunk identifier is greater than the previous access count for the hottest chunk. In this case, a hottest chunk logic for identifying the chunk identifier as the hottest chunk is further included.

As a twenty-third phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the twenty-first phrase, and the node entry further includes the access count for the chunk identifier and the expiration time for the chunk identifier. do.

As a twenty-fourth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the sixteenth phrase, and the degrading logic comprises: a comparator for determining whether an expiration time for the chunk identifier has passed; And a reducer for reducing the stream identifier if the expiration time for the chunk identifier has elapsed, and the second line up unit responds to the reduced stream identifier for the chunk identifier, and the plurality of stream identifiers It operates to place the chunk identifier in a second of the corresponding plurality of queues.

As the twenty-fifth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the twenty-fourth phrase, and the degrading logic is based on the chunk identifier when the chunk identifier is at the front of the plurality of queues. It operates to determine when the stream identifier is degraded.

As the twenty-sixth phrase, an embodiment of the technical idea of the present invention includes the solid state drive according to the twenty-fourth phrase, and the degrading logic is the chunk identifier when the chunk identifier is at the front of the first of a plurality of queues. It operates to determine when the stream identifier is degraded based on.

As the 27th phrase, an embodiment of the technical idea of the present invention includes a driver for use in a computing system, and the driver includes a solid state drive (SSD) including a logical block address (LBA). A receiver for receiving a write command for ); A logical block address mapper for mapping the logical block exchange to a chunk identifier (ID); Stream selection logic for selecting a stream identifier based on the chunk identifier using a chunk-stream mapper stored in a memory of the host computer system; A stream identifier adder for adding the stream identifier to the write command; A line-up device for placing the chunk identifier in a submission queue stored in the memory; And background logic to remove the chunk identifier from the submission queue and update the chunk-stream mapper.

As a twenty-eighth phrase, an embodiment of the technical idea of the present invention includes a driver according to the twenty-seventh phrase, and further includes a transmitter that transmits the write command to the solid state drive using the stream identifier.

As the 29th phrase, an embodiment of the technical idea of the present invention includes the driver according to the 27th phrase, and the logical block address mapper includes an address mask for determining the chunk identifier by covering a part of the logical block address.

As a thirtieth phrase, an embodiment of the technical idea of the present invention includes the driver according to the twenty-seventh phrase, and the logical block address mapper divides the logical block address by a chunk size and identifies the chunk. arithmetic logic unit).

As the 31st phrase, an embodiment of the technical idea of the present invention includes the driver according to the 27th phrase, and the chunk-stream mapper includes a sequential, frequency, and latest (SFR, Sequential, Frequency, Recency) table, and the The SFR table includes the chunk identifier and the stream identifier for the chunk identifier.

As the 32nd phrase, an embodiment of the technical idea of the present invention includes the driver according to the 31st phrase, and the background logic selects a previous stream if the logical block address is sequential to the second logical block address of the previous write command. Includes sequential logic.

As a 33rd phrase, an embodiment of the technical idea of the present invention includes the driver according to the 32nd phrase, and the previous write command is in a window preceding the write command having a window size.

As phrase 34, an embodiment of the technical idea of the present invention includes a driver according to phrase 32, wherein the window size is the number of cores of the processor of the host computer system and software executed by the processor of the host computer system. It is determined in response to at least one of the number of sources.

As a thirty-fifth phrase, an embodiment of the technical idea of the present invention includes the driver according to the thirty-first phrase, and the background logic includes a current access time for the chunk identifier, a previous access time for the chunk identifier, and a decline period. The latest logic to calculate the latest weight based on; An access count adjuster for producing an adjusted access count by adjusting an access count for the chunk identifier based on the latest weight; And a stream identifier adjuster for adjusting the stream identifier based on the adjusted access count for the chunk identifier.

As phrase 36, an embodiment of the technical idea of the present invention includes a driver according to phrase 35, and the SFR table includes the current access time for the chunk identifier, the previous access time for the chunk identifier, and the The access count for the chunk identifier is further included.

As the 37th phrase, an embodiment of the technical idea of the present invention includes the driver according to the 35th phrase, and the latest logic calculates the latest weight of 2 (the current access time for the chunk identifier and the current access time for the chunk identifier). The difference between the previous access times is divided by the decay period).

As the 38th phrase, an embodiment of the technical idea of the present invention includes the driver according to the 35th phrase, and the access count adjuster calculates the adjusted access count for the chunk identifier (to the access count for the chunk identifier). 1) divided by the latest weight above.

As phrase 39, an embodiment of the technical idea of the present invention includes the driver according to phrase 35, and the stream identifier controller calculates the stream identifier as a log of the adjusted access count for the chunk identifier.

As phrase 40, an embodiment of the technical idea of the present invention includes the driver according to the 27th phrase, the chunk-stream mapper includes a node entry, and the node entry includes the chunk identifier and the chunk identifier. Contains the stream identifier.

As phrase 41, an embodiment of the technical idea of the present invention includes a driver according to phrase 40, wherein the background logic includes: a promotion logic for determining when to promote the stream identifier based on the chunk identifier; A second stringer for disposing the chunk identifier at the first of a plurality of queues corresponding to a plurality of stream identifiers in response to the stream identifier for the chunk identifier; And a degrading logic for determining when to degrade the stream identifier based on the chunk identifier.

As the 42nd phrase, an embodiment of the technical idea of the present invention includes the driver according to the 41st phrase, and the promotion logic includes: an increaser for increasing an access count for the chunk identifier; And a stream identifier adjuster for determining the stream identifier in response to the access count for the chunk identifier.

As phrase 43, an embodiment of the technical idea of the present invention includes the driver according to phrase 42, and the stream identifier adjuster operates to determine the stream identifier based on a log of the access count for the stream identifier.

As phrase 44, an embodiment of the technical idea of the present invention includes the driver according to phrase 42, and the promotion logic further includes chunk expiration logic for calculating an expiration time for the chunk identifier.

As phrase 45, an embodiment of the technical idea of the present invention includes the driver according to phrase 44, and the chunk expiration logic determines the expiration time for the chunk identifier between the access count for the chunk identifier and the device lifetime. It operates to operate as a sum.

As clause 46, an embodiment of the technical idea of the present invention includes the driver according to clause 44, and the device lifetime is a difference between a last access time for the hottest chunk and a previous access time for the hottest chunk. .

As the 47th phrase, an embodiment of the technical idea of the present invention includes the driver according to the 46th phrase, and the promotion logic is configured when the access count for the chunk identifier is greater than the previous access count for the hottest chunk. It further includes hottest chunk logic for identifying the chunk identifier as the hottest chunk.

As phrase 48, an embodiment of the technical idea of the present invention includes the driver according to phrase 46, and the node entry further includes the access count for the chunk identifier and the expiration time for the chunk identifier.

As phrase 49, an embodiment of the technical idea of the present invention includes a driver according to phrase 41, and the degrading logic includes: a comparator for determining whether an expiration time for the chunk identifier has passed; And a reducer for reducing the stream identifier if the expiration time for the chunk identifier has elapsed, and the second line up unit responds to the reduced stream identifier for the chunk identifier, and the plurality of stream identifiers It operates to place the chunk identifier in a second of the corresponding plurality of queues.

As phrase 50, an embodiment of the technical idea of the present invention includes the driver according to phrase 49, and the degrading logic is based on the chunk identifier when the chunk identifier is at the front of the plurality of queues. It operates to determine when to downgrade.

As phrase 51, an embodiment of the technical idea of the present invention includes the driver according to phrase 49, and the demotion logic is based on the chunk identifier when the chunk identifier is at the front of the first of a plurality of queues. Thus, it is operated to determine when the stream identifier is degraded.

As phrase 52, an embodiment of the technical idea of the present invention includes a method. The method includes receiving a write command from a software source; Determining a logical block address (LBA) in the write command; Identifying a chunk identifier (ID) for a chunk in a solid state drive (SSD) including the logical block address; Accessing a stream identifier associated with the chunk identifier; Allocating the stream identifier to the write command; Processing the write command in the solid state drive using the allocated stream identifier; And performing a background update of the stream identifier associated with the chunk identifier.

As phrase 53, an embodiment of the technical idea of the present invention includes the method of phrase 52, which is implemented in one of a file system layer, a block layer, or a device driver layer of the host computer system.

As phrase 54, an embodiment of the technical idea of the present invention includes the method of phrase 52, which is implemented in the flash conversion layer of the solid state drive.

As phrase 55, an embodiment of the technical idea of the present invention includes the method of phrase 52, wherein the step of identifying a chunk identifier for a chunk in the solid state drive including the logical block address includes an address from the logical block address. And identifying the chunk identifier using a mask.

As phrase 56, an embodiment of the technical idea of the present invention includes the method of phrase 52, wherein the step of identifying a chunk identifier for a chunk in the solid state drive including the logical block address includes the logical block address. Dividing by the number of sectors in the chunk.

As phrase 57, an embodiment of the technical idea of the present invention includes the method of phrase 52, and the step of allocating the stream identifier to the write command includes adding the stream identifier to the write command as a tag. do.

As phrase 58, an embodiment of the technical idea of the present invention includes the method of phrase 52, comprising the steps of: determining whether the logical block address is sequential to a second logical block address of a second write command; And determining the second stream identifier assigned to the second write command if the logical block address is sequential to the second logical block address of the second write command. And allocating the second stream identifier to the write command.

As the 59th phrase, an embodiment of the technical idea of the present invention includes the method of the 58th phrase, and the second write command is in a window preceding the write command.

As phrase 60, an embodiment of the technical idea of the present invention includes the method of phrase 59, and further includes identifying the window.

As phrase 61, an embodiment of the technical idea of the present invention includes the method of phrase 60, and the step of identifying the window includes the number of cores of the processor of the host computer system including the solid state drive and the solid state. Identifying a window size for the window in response to at least one of the number of software sources executing on the processor of the host computer system including a drive.

As the 62nd phrase, an embodiment of the technical idea of the present invention includes the method of the 59th phrase, including the steps of identifying the oldest writing command in the window; And replacing the oldest write command in the window with the write command.

As phrase 63, an embodiment of the technical idea of the present invention includes the method of phrase 52, wherein the performing background update of the stream identifier associated with the chunk identifier includes: adding the chunk identifier to a submission queue; And removing the chunk identifier from the submission queue when the chunk identifier is at the forefront of the submission queue.

As phrase 64, an embodiment of the technical idea of the present invention includes the method of phrase 52, wherein performing background update of the stream identifier associated with the chunk identifier includes increasing an access count for the chunk identifier. ; Calculating a latest weight for the chunk identifier in response to a current access time and a previous access time for the chunk identifier; Updating the access count for the chunk identifier in response to the latest weighting; And determining the stream identifier for the chunk identifier in response to the updated access count.

As phrase 65, an embodiment of the technical idea of the present invention includes the method of phrase 64, and calculating the latest weight for the chunk identifier in response to a current access time and a previous access time for the chunk identifier And calculating the latest weight as the power of 2 (dividing the difference between the current access time and the previous access time for the chunk identifier by a decay period).

As phrase 66, an embodiment of the technical idea of the present invention includes the method of phrase 65, wherein the updating of the access count for the chunk identifier in response to the latest weighting comprises changing the access count to the latest weighting. Includes the step of dividing.

As phrase 67, an embodiment of the technical idea of the present invention includes the method of phrase 64, and determining the stream identifier for the chunk identifier in response to the updated access count comprises: And calculating the stream identifier as a log of the produced access count.

As phrase 68, an embodiment of the technical idea of the present invention includes the method of phrase 52, and in the step of performing a background update of the stream identifier associated with the chunk identifier, the chunk identifier is one of a plurality of queues. Placing in a queue corresponding to the stream identifier; And determining whether to degrade the chunk identifier when the chunk identifier reaches the front of the queue.

As phrase 69, an embodiment of the technical idea of the present invention includes the method of phrase 68, and the step of arranging the chunk identifier in a queue corresponding to the stream identifier among a plurality of queues includes access to the chunk identifier. Increasing the count; And determining the stream identifier for the chunk identifier in response to the access count for the chunk identifier.

As phrase 70, an embodiment of the technical idea of the present invention includes the method of phrase 69, wherein determining the stream identifier for the chunk identifier in response to the access count for the chunk identifier comprises the chunk identifier And calculating the stream identifier for the as a log of the access count for the chunk identifier.

As the 71st phrase, an embodiment of the technical idea of the present invention includes the method of the 69th phrase, and the step of arranging the chunk identifier in a queue corresponding to the stream identifier among a plurality of queues includes the And if the access count exceeds the second access count for the hottest chunk, identifying the chunk identifier as a new hottest chunk.

As phrase 72, an embodiment of the technical idea of the present invention includes the method of phrase 71, wherein the step of identifying the chunk identifier as a new hottest chunk includes determining a device lifetime as a current access time for the chunk identifier and the chunk. And calculating the difference between the previous access times for the identifier.

As phrase 73, an embodiment of the technical idea of the present invention includes the method of phrase 68, wherein the performing background update of the stream identifier associated with the chunk identifier comprises the chunk identifier in response to the access count and the device lifetime. Determining an expiration time for the identifier, and determining whether to degrade the chunk identifier when the chunk identifier reaches the front of the queue, if the expiration time for the chunk identifier has passed: the stream Removing the chunk identifier from the queue corresponding to the identifier; Reducing the stream identifier; And placing the chunk identifier in a second queue corresponding to the reduced stream identifier.

As phrase 74, an embodiment of the technical idea of the present invention includes the method of phrase 73, and determining an expiration time for the chunk identifier in response to the access count and the device lifetime is the hottest device lifetime. And calculating the difference between the last access time for the chunk and the previous access time for the hottest chunk.

As phrase 75, an embodiment of the technical idea of the present invention includes the method of phrase 73, wherein when the chunk identifier reaches the front of the queue, determining whether to downgrade the chunk identifier includes: If the expiration time has passed and if the chunk identifier is the hottest chunk, selecting a second chunk identifier of the queue corresponding to the stream identifier as a new hottest chunk.

As phrase 76, an embodiment of the technical idea of the present invention includes an article including a non-transitory storage medium. The non-transitory storage medium includes instructions stored thereon, and when the instructions are executed by a machine: receiving a write instruction from a software source; Determining a logical block address (LBA) in the write command; Identifying a chunk identifier (ID) for a chunk in a solid state drive (SSD) including the logical block address; Accessing a stream identifier associated with the chunk identifier; Allocating the stream identifier to the write command; Processing the write command in the solid state drive using the allocated stream identifier; In addition, a step of performing a background update of the stream identifier associated with the chunk identifier is triggered.

As phrase 77, an embodiment of the technical idea of the present invention includes the article of phrase 76, and the method is implemented in one of a file system layer, a block layer, or a device driver layer of the host computer system.

As clause 78, an embodiment of the technical idea of the present invention includes the article of clause 76, and the method is implemented in the flash conversion layer of the solid state drive.

As phrase 79, an embodiment of the technical idea of the present invention includes the article of phrase 76, and the step of identifying a chunk identifier for a chunk in the solid state drive including the logical block address includes an address from the logical block address. And identifying the chunk identifier using a mask.

As phrase 80, an embodiment of the technical idea of the present invention includes the article of phrase 76, and the step of identifying a chunk identifier for a chunk in the solid state drive including the logical block address includes the logical block address. Dividing by the number of sectors in the chunk.

As the 81st phrase, an embodiment of the technical idea of the present invention includes the article of the 76th phrase, and the step of allocating the stream identifier to the write command includes adding the stream identifier to the write command as a tag. do.

As phrase 82, an embodiment of the technical idea of the present invention includes the article of phrase 76, and determining whether the logical block address is sequential to a second logical block address of a second write command; And determining the second stream identifier assigned to the second write command if the logical block address is sequential to the second logical block address of the second write command. And allocating the second stream identifier to the write command.

As the 83rd phrase, an embodiment of the technical idea of the present invention includes the article of the 82nd phrase, and the second write command is in a window preceding the write command.

As phrase 84, an embodiment of the technical idea of the present invention includes the article of phrase 83, and when the non-transitory storage medium is executed by the machine: further includes instructions for causing the step of identifying the window. do.

As phrase 85, an embodiment of the technical idea of the present invention includes the article of clause 84, and the step of identifying the window includes the number of cores of the processor of the host computer system including the solid state drive and the solid state. Identifying a window size for the window in response to at least one of the number of software sources executing on the processor of the host computer system including a drive.

As the 86th phrase, an embodiment of the technical idea of the present invention includes the article of the 83rd phrase, and when the non-temporary storage medium is executed by the machine: identifying the oldest write command in the window; And instructions for causing the step of replacing the oldest write command with the write command in the window.

As the 87th phrase, an embodiment of the technical idea of the present invention includes the article of the 76th phrase, and the performing background update of the stream identifier associated with the chunk identifier includes: adding the chunk identifier to a submission queue; And removing the chunk identifier from the submission queue when the chunk identifier is at the head of the submission queue.

As phrase 88, an embodiment of the technical idea of the present invention includes the article of phrase 76, and performing background update of the stream identifier associated with the chunk identifier includes increasing an access count for the chunk identifier. ; Calculating a latest weight for the chunk identifier in response to a current access time and a previous access time for the chunk identifier; Updating the access count for the chunk identifier in response to the latest weighting; And determining the stream identifier for the chunk identifier in response to the updated access count.

As phrase 89, an embodiment of the technical idea of the present invention includes the article of phrase 88, and calculating the latest weight for the chunk identifier in response to a current access time and a previous access time for the chunk identifier And calculating the latest weight as the power of 2 (dividing the difference between the current access time and the previous access time for the chunk identifier by a decay period).

As the 90th phrase, an embodiment of the technical idea of the present invention includes the article of the 89th phrase, and in response to the latest weighting, updating the access count for the chunk identifier may include changing the access count to the latest weighting. Includes the step of dividing.

As phrase 91, an embodiment of the technical idea of the present invention includes the article of phrase 88, and determining the stream identifier for the chunk identifier in response to the updated access count comprises: And calculating the stream identifier as a log of the produced access count.

As phrase 92, an embodiment of the technical idea of the present invention includes the article of phrase 76, and the performing background update of the stream identifier associated with the chunk identifier includes the chunk identifier from among a plurality of queues. Placing in a queue corresponding to the identifier; And determining whether to downgrade the chunk identifier when the chunk identifier reaches the head of the queue.

As phrase 93, an embodiment of the technical idea of the present invention includes the article of phrase 92, and the step of arranging the chunk identifier in a queue corresponding to the stream identifier among a plurality of queues includes access to the chunk identifier. Increasing the count; And determining the stream identifier for the chunk identifier in response to the access count for the chunk identifier.

As the 94th phrase, an embodiment of the technical idea of the present invention includes the article of the 93rd phrase, and determining the stream identifier for the chunk identifier in response to the access count for the chunk identifier comprises the chunk identifier And calculating the stream identifier for the as a log of the access count for the chunk identifier.

As the 95th phrase, an embodiment of the technical idea of the present invention includes the article of the 93rd phrase, and the step of arranging the chunk identifier in a queue corresponding to the stream identifier among a plurality of queues includes the And if the access count exceeds the second access count for the hottest chunk, identifying the chunk identifier as a new hottest chunk.

As phrase 96, an embodiment of the technical idea of the present invention includes the article of phrase 95, and the step of identifying the chunk identifier as a new hottest chunk includes determining a device lifetime as a current access time for the chunk identifier and the chunk. And calculating the difference between the previous access times for the identifier.

As phrase 97, an embodiment of the technical idea of the present invention includes the article of phrase 92, and the performing background update of the stream identifier associated with the chunk identifier comprises the chunk identifier in response to the access count and the device lifetime. Determining an expiration time for the identifier, and determining whether to degrade the chunk identifier when the chunk identifier reaches the front of the queue, if the expiration time for the chunk identifier has passed: the stream Removing the chunk identifier from the queue corresponding to the identifier; Reducing the stream identifier; And placing the chunk identifier in a second queue corresponding to the reduced stream identifier.

As phrase 98, an embodiment of the technical idea of the present invention includes the article of phrase 97, and determining an expiration time for the chunk identifier in response to the access count and the device lifetime is the hottest device lifetime. And calculating the difference between the last access time for the chunk and the previous access time for the hottest chunk.

As the 99th phrase, an embodiment of the technical idea of the present invention includes the article of the 97th phrase, and when the chunk identifier reaches the head of the queue, determining whether to downgrade the chunk identifier includes: If the expiration time has passed and if the chunk identifier is the hottest chunk, selecting a second chunk identifier of the queue corresponding to the stream identifier as a new hottest chunk.

As a result, in view of a wide and various permutations of the described embodiments, the detailed description and the accompanying materials are merely exemplary, and do not limit the scope of the technical idea of the present invention. Accordingly, as in the appended claims, all such changes belong to the scope of the technical spirit of the claims to be described below and equivalent thereto.

105: machine 110: processor
115: memory 120: solid state drive
205: clock 210: network connector
215: bus 220: user interface
225: input/output engine 305: host interface logic
310: solid state drive controller 315-1~315-8: flash chips
320-1 to 320-4: channels 325: flash conversion layer
330: storage 335: submission queue
340: chunk-stream mapper 405: receiver
410: logical block address mapper 415: stream selection logic
420: stream identifier adder 425: transmitter
430: line up 435: background logic
905: sequential logic 910: modern logic
915: access count adjuster 920: stream identifier adjuster
1005: storage 1010: comparator

Claims (20)

For solid state drives (SSDs):
A flash memory for storing data;
Support means for a plurality of device streams in the solid state drive;
A solid state controller that includes storage for a submission queue and a chunk-stream mapper, and manages writing data to the flash memory in response to a plurality of write commands; And
A flash transform layer, wherein the flash transform layer is:
A receiver for receiving a write command including a logical block address (LBA);
A logical block address mapper for mapping the logical block address to a chunk identifier (ID);
Stream selection logic for selecting a stream identifier based on the chunk identifier using the chunk-stream mapper;
A stream identifier adder for adding the stream identifier to the write command;
A queue for placing the chunk identifier in the submission queue; And
Solid state drive comprising background logic to remove the chunk identifier from the submission queue and update the chunk-stream mapper. The method of claim 1,
The chunk-stream mapper includes a sequential, frequency, and latest (SFR, Sequential, Frequency, Recency) table, and the SFR table includes the chunk identifier and the stream identifier for the chunk identifier. The method of claim 2,
The background logic includes sequential logic for selecting a previous stream if the logical block address is sequential to a second logical block address of a previous write command. The method of claim 2,
The background logic is:
Latest logic for calculating a latest weight based on a current access time for the chunk identifier, a previous access time for the chunk identifier, and a decay period;
An access count adjuster for producing an adjusted access count by adjusting an access count for the chunk identifier based on the latest weight; And
And a stream identifier adjuster for adjusting the stream identifier based on the adjusted access count for the chunk identifier. The method of claim 1,
The chunk-stream mapper comprises a node entry, the node entry comprising the chunk identifier and the stream identifier for the chunk identifier. The method of claim 5,
The background logic is:
Promotion logic for determining when to promote the stream identifier based on the chunk identifier;
A second stringer for disposing the chunk identifier at first of a plurality of queues corresponding to a plurality of stream identifiers in response to the stream identifier for the chunk identifier; And
And a degrading logic for determining when to degrade the stream identifier based on the chunk identifier. The method of claim 6,
The degrading logic is:
A comparator for determining whether an expiration time for the chunk identifier has passed; And
If the expiration time for the chunk identifier has passed, a reducer for reducing the stream identifier,
The second stringer is operative to place the chunk identifier in a second of the plurality of queues corresponding to the plurality of stream identifiers in response to the reduced stream identifier for the chunk identifier. A non-transitory storage medium storing instructions that, when executed by a machine, cause:
A receiver for receiving a write command for a solid state drive (SSD) including a logical block address (LBA);
A logical block address mapper for mapping the logical block address to a chunk identifier (ID);
Stream selection logic for selecting a stream identifier based on the chunk identifier using a chunk-stream mapper stored in a memory of the host computer system;
A stream identifier adder for adding the stream identifier to the write command;
A line-up device for placing the chunk identifier in a submission queue stored in the memory; And
A non-transitory storage medium comprising background logic to remove the chunk identifier from the submission queue and update the chunk-stream mapper. The method of claim 8,
The chunk-stream mapper includes a sequential, frequency, and latest (SFR, Sequential, Frequency, Recency) table, and the SFR table includes the chunk identifier and the stream identifier for the chunk identifier. The method of claim 9,
The background logic includes sequential logic for selecting a previous stream if the logical block address is sequential to a second logical block address of a previous write command. The method of claim 9,
The background logic is:
Latest logic for calculating a latest weight based on a current access time for the chunk identifier, a previous access time for the chunk identifier, and a decay period;
An access count adjuster for producing an adjusted access count by adjusting an access count for the chunk identifier based on the latest weight; And
And a stream identifier adjuster for adjusting the stream identifier based on the adjusted access count for the chunk identifier. The method of claim 8,
The chunk-stream mapper includes a node entry, and the node entry includes the chunk identifier and the stream identifier for the chunk identifier. The method of claim 12,
The background logic is:
Promotion logic for determining when to promote the stream identifier based on the chunk identifier;
A second stringer for disposing the chunk identifier at first of a plurality of queues corresponding to a plurality of stream identifiers in response to the stream identifier for the chunk identifier; And
A non-transitory storage medium comprising a demotion logic for determining when to degrade the stream identifier based on the chunk identifier. The method of claim 13,
The degrading logic is:
A comparator for determining whether an expiration time for the chunk identifier has passed; And
If the expiration time for the chunk identifier has passed, a reducer for reducing the stream identifier,
The second line-up device is a non-transitory storage medium operable to place the chunk identifier in a second of the plurality of queues corresponding to the plurality of stream identifiers in response to the reduced stream identifier for the chunk identifier. . Receiving a write command from a software source;
Determining a logical block address (LBA) in the write command;
Identifying a chunk identifier (ID) for a chunk in a solid state drive (SSD) including the logical block address;
Accessing a stream identifier associated with the chunk identifier;
Allocating the stream identifier to the write command;
Processing the write command in the solid state drive using the allocated stream identifier; And
And performing a background update of the stream identifier associated with the chunk identifier. The method of claim 15,
Determining whether the logical block address is sequential to a second logical block address of a second write command; And
If the logical block address is sequential to the second logical block address of the second write command:
Determining a second stream identifier assigned to the second write command; And
Allocating the second stream identifier to the write command. The method of claim 15,
Performing a background update of the stream identifier associated with the chunk identifier comprises:
Adding the chunk identifier to a submission queue; And
Removing the chunk identifier from the submission queue when the chunk identifier is at the forefront of the submission queue. The method of claim 15,
Performing a background update of the stream identifier associated with the chunk identifier comprises:
Increasing an access count for the chunk identifier;
Calculating a latest weight for the chunk identifier in response to a current access time and a previous access time for the chunk identifier;
Updating the access count for the chunk identifier in response to the latest weighting; And
Determining the stream identifier for the chunk identifier in response to the updated access count. The method of claim 15,
Performing a background update of the stream identifier associated with the chunk identifier comprises:
Placing the chunk identifier in a queue corresponding to the stream identifier, which is one of a plurality of queues; And
Determining whether to degrade the chunk identifier when the chunk identifier reaches the front of the queue. The method of claim 19,
Performing background update of the stream identifier associated with the chunk identifier includes determining an expiration time for the chunk identifier in response to an access count and a device lifetime; And
The step of determining whether to degrade the chunk identifier when the chunk identifier reaches the front of the queue is if the expiration time for the chunk identifier has passed:
Removing the chunk identifier from the queue corresponding to the stream identifier;
Reducing the stream identifier; And
Placing the chunk identifier in a second queue corresponding to the reduced stream identifier.

Images